Intrinsic cognitive models

Theories concerning the structure, or format, of mental representation should (1) be formulated in mechanistic, rather than metaphorical terms; (2) do justice to several philosophical intuitions about mental representation; and (3) explain the human capacity to predict the consequences of worldly alterations (i.e., to think before we act). The hypothesis that thinking involves the application of syntax‐sensitive inference rules to syntactically structured mental representations has been said to satisfy all three conditions. An alternative hypothesis is that thinking requires the construction and manipulation of the cognitive equivalent of scale models. A reading of this hypothesis is provided that satisfies condition (1) and which, even though it may not fully satisfy condition (2), turns out (in light of the frame problem) to be the only known way to satisfy condition (3).     

Symbolically speaking: a connectionist model of sentence production

The ability to combine words into novel sentences has been used to argue that humans have symbolic language production abilities. Critiques of connectionist models of language often center on the inability of these models to generalize symbolically (Fodor & Pylyshyn, 1988; Marcus, 1998). To address these issues, a connectionist model of sentence production was developed. The model had variables (role‐concept bindings) that were inspired by spatial representations (Landau & Jackendoff, 1993). In order to take advantage of these variables, a novel dual‐pathway architecture with event semantics is proposed and shown to be better at symbolic generalization than several variants. This architecture has one pathway for mapping message content to words and a separate pathway that enforces sequencing constraints. Analysis of the model's hidden units demonstrated that the model learned different types of information in each pathway, and that the model's compositional behavior arose from the combination of these two pathways. The model's ability to balance symbolic and statistical behavior in syntax acquisition and to model aphasic double dissociations provided independent support for the dual‐pathway architecture.     

The role of cognitive architectures in general artificial intelligence

The term “Cognitive Architectures” indicates both abstract models of cognition, in natural and artificial agents, and the software instantiations of such models which are then employed in the field of Artificial Intelligence (AI). The main role of Cognitive Architectures in AI is that one of enabling the realization of artificial systems able to exhibit intelligent behavior in a general setting through a detailed analogy with the constitutive and developmental functioning and mechanisms underlying human cognition. We provide a brief overview of the status quo and the potential role that Cognitive Architectures may serve in the fields of Computational Cognitive Science and Artificial Intelligence (AI) research.     

A model of plausibility

Plausibility has been implicated as playing a critical role in many cognitive phenomena from comprehension to problem solving. Yet, across cognitive science, plausibility is usually treated as an operationalized variable or metric rather than being explained or studied in itself. This article describes a new cognitive model of plausibility, the Plausibility Analysis Model (PAM), which is aimed at modeling human plausibility judgment. This model uses commonsense knowledge of concept-coherence to determine the degree of plausibility of a target scenario. In essence, a highly plausible scenario is one that fits prior knowledge well: with many different sources of corroboration, without complexity of explanation, and with minimal conjecture. A detailed simulation of empirical plausibility findings is reported, which shows a close correspondence between the model and human judgments. In addition, a sensitivity analysis demonstrates that PAM is robust in its operations.     

Modeling recognition memory using the similarity structure of natural input

The natural input memory (NIM) model is a new model for recognition memory that operates on natural visual input. A biologically informed perceptual preprocessing method takes local samples (eye fixations) from a natural image and translates these into a feature-vector representation. During recognition, the model compares incoming preprocessed natural input to stored representations. By complementing the recognition memory process with a perceptual front end, the NIM model is able to make predictions about memorability based directly on individual natural stimuli. We demonstrate that the NIM model is able to simulate experimentally obtained similarity ratings and recognition memory for individual stimuli (i.e., face images).     

Modeling how, when, and what is learned in a simple fault-finding task

We have developed a process model that learns in multiple ways while finding faults in a simple control panel device. The model predicts human participants' learning through its own learning. The model's performance was systematically compared to human learning data, including the time course and specific sequence of learned behaviors. These comparisons show that the model accounts very well for measures such as problem-solving strategy, the relative difficulty of faults, and average fault-finding time. More important, because the model learns and transfers its learning across problems, it also accounts for the faster problem-solving times due to learning when examined across participants, across faults, and across the series of 20 trials on an individual participant basis. The model shows how learning while problem solving can lead to more recognition-based performance, and helps explain how the shape of the learning curve can arise through learning and be modified by differential transfer. Overall, the quality of the correspondence appears to have arisen from procedural, declarative, and episodic learning all taking place within individual problem-solving episodes.     

语义问题与人工智能模型构造的系统观点

基于"物理符号系统假设"的传统人工智能采信了低阶结构不连续的思想,将概念化与概念的语义基础分离,把对思维过程的模拟看作是能用形式化方法来实现的。但事实证明,这种后验性方法存在理论和实现上的双重危机,要完成有实际意义的、有创新性的智能行为丰富的语义是必需的,这使语义问题成为人工智能不同应用分支中的焦点,它包括语义获取、表达和使用三个方面。而要实现对语义问题的认识和解决就必须把它和其它智能行为作为一个连续的、相关的、不可分割的认知结构进行完整地考察,以系统的观点来看待智能模型的构造问题。这个认知结构统一性的基石就是基于神经生理基础的、以知觉的心理生理学解释为依据的、对语义的直接表达,这为统一以神经系统动力学为模型的其它各种智能行为提供了基础。     

A real‐world rational agent: unifying old and new AI

Explanations of cognitive processes provided by traditional artificial intelligence were based on the notion of the knowledge level. This perspective has been challenged by new AI that proposes an approach based on embodied systems that interact with the real‐world. We demonstrate that these two views can be unified. Our argument is based on the assumption that knowledge level explanations can be defined in the context of Bayesian theory while the goals of new AI are captured by using a well established robot based model of learning and problem solving, called Distributed Adaptive Control (DAC). In our analysis we consider random foraging and we prove that minor modifications of the DAC architecture renders a model that is equivalent to a Bayesian analysis of this task. Subsequently, we compare this enhanced, “rational,” model to its “non‐rational” predecessor and a further control condition using both simulated and real robots, in a variety of environments. Our results show that the changes made to the DAC architecture, in order to unify the perspectives of old and new AI, also lead to a significant improvement in random foraging.     

An organizing model for recent cognitive science work on the self

An organizing model of ‘the self’ emerges from applying various kinds of brain injury to recent cognitive science and philosophical work on ‘the self’. This model unifies various contents and mechanisms central to current notions of the self. The article then highlights several criteria and aspects of this notion of self. Qualities of the right type and level of psychological significance delineate ‘the self’ as an organizing concept useful for recent philosophical work and cognitive science research.     

Toward ethical cognitive architectures for the development of artificial moral agents

New technologies based on artificial agents promise to change the next generation of autonomous systems and therefore our interaction with them. Systems based on artificial agents such as self-driving cars and social robots are examples of this technology that is seeking to improve the quality of people’s life. Cognitive architectures aim to create some of the most challenging artificial agents commonly known as bio-inspired cognitive agents. This type of artificial agent seeks to embody human-like intelligence in order to operate and solve problems in the real world as humans do. Moreover, some cognitive architectures such as Soar, LIDA, ACT-R, and iCub try to be fundamental architectures for the Artificial General Intelligence model of human cognition. Therefore, researchers in the machine ethics field face ethical questions related to what mechanisms an artificial agent must have for making moral decisions in order to ensure that their actions are always ethically right. This paper aims to identify some challenges that researchers need to solve in order to create ethical cognitive architectures. These cognitive architectures are characterized by the capacity to endow artificial agents with appropriate mechanisms to exhibit explicit ethical behavior. Additionally, we offer some reasons to develop ethical cognitive architectures. We hope that this study can be useful to guide future research on ethical cognitive architectures.     

Face recognition algorithms and the other‐race effect: computational mechanisms for a developmental contact hypothesis

People recognize faces of their own race more accurately than faces of other races. The “contact” hypothesis suggests that this “other‐race effect” occurs as a result of the greater experience we have with own‐ versus other‐race faces. The computational mechanisms that may underlie different versions of the contact hypothesis were explored in this study. We replicated the other‐race effect with human participants and evaluated four classes of computational face recognition algorithms for the presence of an other‐race effect. Consistent with the predictions of a developmental contact hypothesis, “experience‐based models” demonstrated an other‐race effect only when the representational system was developed through experience that warped the perceptual space in a way that was sensitive to the overall structure of the model's experience with faces of different races. When the model's representation relied on a feature set optimized to encode the information in the learned faces, experience‐based algorithms recognized minority‐race faces more accurately than majority‐race faces. The results suggest a developmental learning process that warps the perceptual space to enhance the encoding of distinctions relevant for own‐race faces. This feature space limits the quality of face representations for other‐race faces.     

The meaningful-based cognitive architecture model of schizophrenia

In subsymbolic operation of the Meaningful-Based Cognitive Architecture (MBCA) the input sensory vector is propagated through a hierarchy of Hopfield-like Network (HLN) functional groups, is recognized and may associatively trigger in the instinctual core goals module as well as in groups of HLNs arranged as pre-causal and pattern memory, vectors propagated to the output motor group of HLNs which produce an output signal. In full causal symbolic operation, the processed sensory input vector is also propagated to the logic/working memory groups of HLNs, where it can be compared to other vectors in the logic/working memory, and produce various outputs in response. The processed sensory input vector can trigger in the instinctual core goals module intuitive logic, intuitive physics, intuitive psychology and intuitive planning procedural vectors, as well as trigger in the causal group of HLNs learned logic, physics, psychology and planning procedural vectors which are also sent to the logic/working memory groups of HLNs. These circuits can allow the MBCA to act causally on information it has never seen before. An example is given of a Python simulation where the MBCA which is controlling a legged robot causally determines that a shallow whitewater river will cause water damage to itself, while if the MBCA is acting associatively only and never having seen whitewater before and normally crossing shallow rivers, will cross the whitewater river and become damaged. While the MBCA does not attempt to replicate biological systems at the neuronal spiking level, its HLNs and the organization of its HLNs are indeed inspired by biological mammalian minicolumns and mammalian brains. The MBCA model leads to the hypothesis that in the course of hominin evolution, HLNs became co-opted into groups of HLNs providing more extensive working memories with causal abilities, unlike non-hominins. While such co-option of the minicolumns can allow advantageous causal symbolic processing integrated with subsymbolic processing, the order of magnitude of increased complexity required for such organization and operation, created a vulnerability in the human brain to psychosis, which does not occur with significant prevalence in non-humans.     

Extending SME to Handle Large‐Scale Cognitive Modeling

Analogy and similarity are central phenomena in human cognition, involved in processes ranging from visual perception to conceptual change. To capture this centrality requires that a model of comparison must be able to integrate with other processes and handle the size and complexity of the representations required by the tasks being modeled. This paper describes extensions to Structure‐Mapping Engine (SME) since its inception in 1986 that have increased its scope of operation. We first review the basic SME algorithm, describe psychological evidence for SME as a process model, and summarize its role in simulating similarity‐based retrieval and generalization. Then we describe five techniques now incorporated into the SME that have enabled it to tackle large‐scale modeling tasks: (a) Greedy merging rapidly constructs one or more best interpretations of a match in polynomial time: O(n²log(n)); (b) Incremental operation enables mappings to be extended as new information is retrieved or derived about the base or target, to model situations where information in a task is updated over time; (c) Ubiquitous predicates model the varying degrees to which items may suggest alignment; (d) Structural evaluation of analogical inferences models aspects of plausibility judgments; (e) Match filters enable large‐scale task models to communicate constraints to SME to influence the mapping process. We illustrate via examples from published studies how these enable it to capture a broader range of psychological phenomena than before.     

Causal cognitive architecture 1: Integration of connectionist elements into a navigation-based framework

The brain-inspired Causal Cognitive Architecture 1 (CCA1) tightly integrates the sensory processing capabilities found in neural networks with many of the causal abilities found in human cognition. Causality emerges not from a central controlling stored program but directly from the architecture. Sensory input vectors are processed by robust association circuitry and then propagated to a navigational temporary map. Instinctive and learned objects and procedures are applied to the same temporary map, with a resultant navigation signal obtained. Navigation can similarly be for the physical world as well as for a landscape of higher cognitive concepts. There is good explainability for causal decisions. A simulation of the CCA1 controlling a search and rescue robot is presented with the goal of finding and rescuing a lost hiker within a grid world. A simulation of the CCA1 controlling a repair robot is presented that can predict the movement of a series of gears.     

Tacit and accessible understanding of language

The empirical nature of our understanding of language is explored. I first show that there are several important and different distinctions between tacit and accessible awareness. I then present empirical evidence concerning our understanding of language. The data suggests that our awareness of sentence-meanings is sometimes merely tacit according to one of these distinctions, but is accessible according to another. I present and defend an interpretation of this mixed view. The present project is shown to impact on several diverse areas, including inferential role semantics and holism, the nature of learning, and the role of linguistics in the law. I am indebted to a number of people for their useful feedback, especially Peter Ludlow, Paul Pietroski, and two anonymous reviewers. Earlier versions of this paper were presented at an Eastern meeting of the APA, a meeting of the Society for Exact Philosophy at Simon Fraser University, and at a semantics workshop in Ottawa, Canada. I greatly appreciate the comments from those audiences.     

Constructive criticism: An evaluation of Buller and Hardcastle's genetic and neuroscientific arguments against Evolutionary Psychology

David Buller and Valerie Hardcastle have argued that various discoveries about the genetics and nature of brain development show that most “central” psychological mechanisms cannot be adaptations because the nature of the contribution from the environment on which they are based shows they are not heritable. Some philosophers and scientists have argued that a strong role for the environment is compatible with high heritability as long as the environment is highly stable down lineages. In this paper I support this view by arguing that the discoveries Buller and Hardcastle refer to either do not show as strong a role for the environment as they suggest, or these discoveries show that the brain's developmental process depends in many cases on input from the environment that is highly stable across generations.     

A Quantitative Empirical Analysis of the Abstract/Concrete Distinction

This study presents original evidence that abstract and concrete concepts are organized and represented differently in the mind, based on analyses of thousands of concepts in publicly available data sets and computational resources. First, we show that abstract and concrete concepts have differing patterns of association with other concepts. Second, we test recent hypotheses that abstract concepts are organized according to association, whereas concrete concepts are organized according to (semantic) similarity. Third, we present evidence suggesting that concrete representations are more strongly feature‐based than abstract concepts. We argue that degree of feature‐based structure may fundamentally determine concreteness, and we discuss implications for cognitive and computational models of meaning.     

The extended cognition thesis: Its significance for the philosophy of (cognitive) science

While the extended cognition (EC) thesis has gained more followers in cognitive science and in the philosophy of mind and knowledge, our main goal is to discuss a different area of significance of the EC thesis: its relation to philosophy of science. In this introduction, we outline two major areas: (I) The role of the thesis for issues in the philosophy of cognitive science, such as: How do notions of EC figure in theories or research programs in cognitive science? Which versions of the EC thesis appear, and with which arguments to support them? (II) The potentials and limits of the EC thesis for topics in general philosophy of science, such as: Can naturalism perhaps be further advanced by means of the more recent EC thesis? Can we understand “big science” or laboratory research better by invoking some version of EC? And can the EC thesis help in overcoming the notorious cognitive/social divide in science studies?     

The role of falsification in the development of cognitive architectures: insights from a lakatosian analysis

It has been suggested that the enterprise of developing mechanistic theories of the human cognitive architecture is flawed because the theories produced are not directly falsifiable. Newell attempted to sidestep this criticism by arguing for a Lakatosian model of scientific progress in which cognitive architectures should be understood as theories that develop over time. However, Newell's own candidate cognitive architecture adhered only loosely to Lakatosian principles. This paper reconsiders the role of falsification and the potential utility of Lakatosian principles in the development of cognitive architectures. It is argued that a lack of direct falsifiability need not undermine the scientific development of a cognitive architecture if broadly Lakatosian principles are adopted. Moreover, it is demonstrated that the Lakatosian concepts of positive and negative heuristics for theory development and of general heuristic power offer methods for guiding the development of an architecture and for evaluating the contribution and potential of an architecture's research program.